During the current grant period the physiologic framework underlying typical speech development has been quantified for the major speech systems (i.e., respiratory, phonatory, and articulatory systems) in children from 3-months to 4-years of age. This proposed continuation integrates these system-specific findings across systems in typically developing children and extends these studies to developmental speech impairment. Well-developed descriptive techniques in speech physiology, phonology, speech and language development, and computational modeling of complex systems provide convergent methods for studying speech development and speech impairment. Neural network-based computational modeling will be used to relate events across speech systems and across levels of observation (i.e., physiologic observations and speech output) and to map the transform of physiologic events to speech output (i.e., phonetic and prosodic). Three hundred 36-48-month-old children, 240 of whom will exhibit speech delay of unknown origin, will be studied using detailed phonetic description, physiologic description (including respiratory, phonatory, and articulatory signals). Connectionist models will identify clustering of covarying physiologic parameters, speech production parameters, and linkages of clusters across these two domains. These analyses will yield a map of the physiologic correlates of speech impairment, and will provide a basis for evaluating current nosologies of child speech delay of unknown origin.